May 14th meeting

Multivariate analysis review

Attendance: Catherine Johnson, Benoit Casault, Emily Chisholm

Meeting minutes

• EC will take notes from today and from BC’s email (May 12) and incorporate them into Reid’s methodology document

  • Some structural reorganization will be done
  • Draft will be reviewed by CJ and BC

• EC and BC will work to update this gitHub repo to include the latest versions of all scripts and data

• Each metric has now been calculated 3 times, once by Dan Reed, then by Reid Steele, then by BC

• BC plotted a comparison of all 3 outputs for comparison

• Plots will be updated to remove jitter, plotting different users with low opacity shapes to distinguish with precision (BC)

• Group went over each plot individually to identify issues

Physical Variables

Physical variable comparison

• AMO

  • DR had used annual values
  • RS and BC used monthly values
  • All three outputs appear to have significant differences
  • EC will investigate differences using other sources for AMO (literature, SART, Res Docs) - BC will send data to facillitate

• Bottom temp 4V includes both 4Vs and 4Vn

• Ice volume

  • BC and RS used updated data source
  • Some small differences between DR older data
  • will maintain newer data set moving forward

• NAO

  • DR used data from Roger Pettipas (pressure)
  • RS and BC used Peter Galbraith’s anomoly data
  • difference in scales makes plot difficult to interpret
  • EC will create scatter plot to compare relationship, eg interannual variability, to ensure data is similar despite scale

• SST

  • difference in data sources between DR and BC/RS
  • RS and BC used Peter Galbraith’s reprocessed data
  • RS combined 4Vs and 4Vn
  • BC used only 4Vs
  • DR analysis was performed before 4V s/n distinction
  • moving forward will use 4Vs

• stratification

  • RS used only HL2 data (error)
  • BC and DR used all SS data
  • EC will ensure all SS data is incorporated in package

• Bottom temperature note

  • 4V and 4W have no 2018 data
  • RS interpolated on BC recommendation using 4X scale value
  • EC will investigate whether this is necessary in PCA
  • CJ will consider best interpolation scheme
  • Incorporate note in future versions of analysis to ensure readers are aware of interpolation

Phenology Variables

Phenology variable comparison

Phenology variable comparison

• Zooplankton peaks
  • BC could not find DR code
  • BC followed methods presented in paper (PCA -> LOESS fit)
  • DR performed a sepearte analysis (PCA) on Calanus stages to get exact timing
  • CJ will look back on DR code for stage abundance PCA
• BioMass Max
  • RS method is wrong, uses abundance not biomass
  • This requires more investigation to confirm
• Cfin I + II + III
  • good agreement between DR and BC
  • BC had bug with 1999 and 2007 values to be resolved
• Cfin HL2
  • agreement between DR and BC
• Note:
  • Dan cleaned up a lot of scripts before leaving, he may have left out some interim steps and data products
  • Ideally would use original DR code, but some cases where BC had to rewrite might be necessary
  • CJ will investigate and look for original code
  • If it can not be found may require follow up with DR
• Ice Area
  • BC and RS used new data provided by Peter Galbraith
  • DR used data from Roger Pettipas
  • will maintain new data source (PG)
• Ice end day
  • All OK
• Bloom parameters
  • some reprocessing of data may have been done in between DR analysis and BC/RS analysis
  • BC will check and follwo up
  • some differences in anomoly calculations between BC and DR
  • may be due to DR lack of reference period in calculations
• SST warming
  • All OK
  • small differences from DR due to processing

Zooplankton variables

Zooplankton Variable Comparisons

Zooplankton Variable Comparisons

Zooplankton Variable Comparisons

Zooplankton Variable Comparisons

Zooplankton Variable Comparisons

Zooplankton Variable Comparisons

Zooplankton Variable Comparisons

• Mostly good agreement between all sources due to least amount of data reprocessing/ adjustment

• Differences in CSS

  • Dr and BC used distinct vars for HL and HL2
  • RS did not use any distinction
  • EC will investigate how this is handled in PCA to ensure consistency

• Metridia

  • Some small differences due to data correction
  • note that pre 2008 ‘Metridia’ maintains zero value, this is likely due to identification methods
  • EC will investigate if Metridia (without species designation) is included in PCA
    • similar issue in Oithona and centropages, less extreme
    • CJ will also investigate raw data

• Reference period

  • BC will double check that the same reference period was used throughout (should be 1999-2015)